RESEARCH ON INTRODUCTION OF ADAPTIVE PREVIEW CONTROL TO ROBOTS

众所周知，利用控制系统输出必须遵循或拒绝的期望信号或干扰信号的未来信息，可以大大提高控制性能。“预览控制”就是其中一种控制方法。在最优调节器理论的基础上，提出了一种用于预瞄控制系统的控制系统综合方法。但是，对于已经设计好的反馈控制系统，预览动作就像前馈动作一样。这样，包含预补偿的总控制系统的鲁棒性就会降低。然后，需要对被控对象参数变化的自适应函数。在本研究中，考虑了以下几点。基于作者提出的“数字加速法”的非线性被控对象的预瞄控制系统设计方法。预览控制系统自适应动作设计方法的发展。机器人机械手的轨迹规划研究。模糊推理、神经网络等学习方法在机器人中的应用。在本研究中，得到以下结果：在数字加速度法的基础上，针对非线性被控对象，提出了不利用被控对象加速度信号的设计新方法。而之前的方法需要加速度信号，这对于一般的被控对象来说是非常困难的。将模糊推理和神经网络引入到预瞄控制系统中，在一定程度上获得了对被控对象参数变化的自适应能力。

It is well known that utilization of future information on desired signal or disturbance signal which the output of the control system must follows or reject makes the control performance considerably improved. "Preview control" is one of such control methods. Control system synthesis method for preview control system has been developed on the basis of optimal regulator theory by the authors. However, the preview action acts like feedfoward action for already designed feedback control system. Then, robustness of the total control system including preview compensation becomes low. Then, adaptive function for parameter variations of the controlled objects is necessary. In this research, the following points are considered.[1]Design method of preview control system for a nonlinear controlled object on the basisi of "digital acceleration method" whichi has been proposed by the authors.[2]Development of design method of adaptive action for preview control system.[4]Study on trajectory planning of the robot manipulator.[4]Sapplication of learning methods such as Fuzzy reasoning and Neural network to robotics.In this research, the following results are obtained.[1]New design method without utilization of the acceleration signal from the controlled object is developed for nonlinear controlled object on the basis of the digital acceleration method. While the previous method needs the acceleration signal which is very hard work for usual controlled object.[2]Adaptation for parameter variations of the controlled object is obtained to some extent by means of introduction of Fuzzy reasoning and Neural network into the preview control system.